Method for operating a disruption tolerant network between an electronic computing device and a terminal by means of a network system

Abstract

Method for operating a fault-resistant network between an electronic computing device and an end device by means of a network system, computer program product, computer-readable storage medium and network system. Method for operating a fault-resistant network (16), comprising the steps of: transmitting data packets (18) from the end device (12) to an electronic computing device (14), wherein each data packet (18) is assigned a respective timestamp (20) and a respective packet number (22) by means of the end device (12); receiving at least a part of the multitude of transmitted data packets (18); determining non-received data packets (18) based on the respective packet number (22); determining at least one time window (24) in which the non-received data packets (18) were sent, based on the timestamps (20) of the received data packets (18);Generating an information message (26) with the at least one determined time window (24); and transmitting the information message (26) to the terminal device (12) for future transmission of data packets (18) via the interference-resistant network (16). The invention further relates to a computer program product, a computer-readable storage medium, and a network system (10).

Description

[0001] Method for operating a fault-resistant network between an electronic computing device and an end device using a network system, computer program product, computer-readable storage medium and network system

[0002] The following invention relates to a method for operating a fault-resistant network system between an electronic computing device and an end device by means of a network system according to claim 1. The invention further relates to a corresponding computer program product, a computer-readable storage medium and a corresponding network system.

[0003] In particular, so-called disruption-tolerant networks (DTN), also known as delay-tolerant networks, represent a class of communication networks designed and developed with unreliable and frequently interrupted connectivity in mind. Examples include mobile ad-hoc networks, local area networks for disaster relief, satellite communication scenarios, and similar applications. These networks tolerate high latency and frequent connection drops and are often highly partitioned. Furthermore, a direct connection between sender and receiver may not exist at any time.In many applications, especially continuous remote sensing tasks, highly restricted end devices, particularly Internet of Things (IoT) devices, must periodically report recorded values ​​back to the network edge to prevent running out of internal memory and for other quality-of-service reasons, such as maintaining the chronological order of the data. If they run out of memory, the time-series values ​​must be discarded. These values ​​are then lost, and the associated time series is often considered invalid if samples are missing.

[0004] Another problem is that these networks typically use a technique called "store and forward" for data transmission, where nodes temporarily store incoming packets and forward them to the next network node as soon as a new connection is established. When a node can no longer store any more packets, a common technique is to, for example, remove the oldest packets from the cache. However, this can lead to packet loss.

[0005] Overall, it is important to use these networks with care, especially as a sender. Constantly flooding these networks with packets is therefore not a beneficial solution.

[0006] From the perspective of IoT devices (especially industrial IoT devices – IIoT) and the network edge, it is therefore of great interest to be able to predict when data packets can next be sent and received, so that they can, for example, switch off their often power-hungry modem and save energy. This is particularly important if the IoT device is powered by non-rechargeable batteries.

[0007] In the prior art, similar problems have been solved in the following way. If the network architecture is known, for example, data links via satellite, simple predictors can be created based on recorded, real-time, and trajectory data, which can then be used to estimate the next connection window. Publicly available data, such as train timetables, can also be used to estimate the next arrival of a data mule, in this case, the train. Furthermore, probabilistic models already exist that estimate the availability, and in particular the position, of a receiver at a specific time.

[0008] The object of the present invention is to provide a method, a computer program product, a computer-readable storage medium and a network system by means of which improved operation of a fault-resistant network between an electronic computing device and an end device can be achieved.

[0009] This problem is solved by a method, a computer program product, a computer-readable storage medium, and a network system according to the independent claims. Advantageous embodiments are specified in the dependent claims.

[0010] One aspect of the invention relates to a method for operating a noise-resistant network between an electronic computer and an end device using a network system. A multitude of data packets are transmitted from the end device to the electronic computer via the noise-resistant network, with each data packet being assigned a respective timestamp and packet number by the end device. At least a portion of the multitude of transmitted data packets is received by the electronic computer. Unreceived data packets are identified by the electronic computer based on their respective packet numbers. At least one time window is determined by the electronic computer, within which the unreceived data packets are sent, based on the timestamps of the received data packets.An information message is generated with the at least determined time window using the electronic computing device and the information message is transmitted to the terminal device for future transmission of data packets via the interference-resistant network using the electronic computing device.

[0011] This allows for improved operation of the interference-resistant network, as it will be possible, for example, to predict when data packet transmission is most efficient. Specifically, it enables the prediction of time windows during which the end device can reliably transmit data packets to the electronic computing unit.

[0012] In particular, the invention describes the IoT device, especially the terminal device, and a corresponding service distance at the network edge. These restricted devices are connected to the service at the network edge via a black-box network, especially the network system. This connection is highly intermittent, and at no time can a direct connection exist between the restricted terminal device and the network edge, for example, in latency-tolerant / interruption-tolerant networks. Due to the "black-box" structure of the network system within the network, a predictive approach is compatible with almost all types of network infrastructures.

[0013] In the proposed approach, the end device uses lookup tables, which are created by the electronic computing unit at the network edge using mathematical models. These lookup tables contain the next estimated start times of connection windows and their duration. This predictive approach is particularly compatible with limited end devices, as all the computing work is performed in the data centers at the network edge, specifically in the electronic computing unit or the network core.

[0014] In particular, the present invention provides that a remote IoT device is connected via an intermittent network (data mules, Subcom, disaster relief network, a combination of all of these, or the like) to a service at the network edge, especially to the electronic computing device. As already mentioned, the network is treated as a so-called black box. For the sake of simplicity, only one IoT device is selected in this setup. However, this approach is naturally also scalable with a large number of additional end devices.

[0015] As a first step, the respective terminal device attempts to download data over the network at regular intervals and therefore sends data packets to the electronic computing unit. These data packets are additionally time-stamped, which specifically describes the local transmission time of the packet. Furthermore, if not already present, an incremental packet ID, specifically the packet number, is appended to the packet. This packet ID is incremented for each packet sent from the terminal device to the electronic computing unit.

[0016] Since the connection between the device and the service at the network edge is only sporadic, some data packets will arrive, while others will not. All received packet numbers are stored, and a so-called window map is created on the electronic computing device's side. This window map represents the intervals in which the terminal device was able to successfully send data packets to the electronic computing device.

[0017] However, due to the often asynchronous nature of interruption-tolerant networks, a connection from the terminal device to the electronic computing device does not automatically mean that a connection from the electronic computing device to the terminal device existed at that time.

[0018] In the next step, the electronic computing device can be trained, based on the large number of connection data records, to predict the next offline / online windows. These predicted windows are then stored in a lookup table, which is transmitted to the end device. The end device can then use this lookup table to send data in the future only when there is a high probability that the network service will receive these packets.

[0019] It should be clarified here that the unreceived data packets are part of the data packets to be transmitted or those already transmitted, and that the non-receipt is therefore due to a transmission error. Thus, the unreceived data packets are transmitted, but not received due to the error. It is desirable, however, that all transmitted data packets are also received. According to an advantageous embodiment, future data packets are transmitted from the terminal device to the electronic computing device within the determined time window. In particular, data packets are only transmitted within the determined time window. This has the particular advantage that, based on the determined time window, it can be ensured that there is a high probability that data packets can be transmitted from the terminal device to the electronic computing device.This prevents data packets from being lost and ensures, for example, that energy-intensive transmission attempts from the end device to the electronic computing unit are carried out. This improves the operation of the interference-resistant network.

[0020] It is further advantageous if the information message is transmitted with an overview table covering at least one determined time window. This overview table corresponds in particular to a lookup table. The overview table is a simple and, in particular, resource-efficient table that can be easily processed by the end device. This requires minimal computing power from the end device, making the procedure particularly advantageous even with very basic end devices.

[0021] In a further advantageous embodiment, the electronic computing device continuously determines time windows and continuously, or at predetermined intervals, adjusts the information message and transmits it to the terminal device. In other words, the electronic computing device can be essentially designed to learn and adapt the overview table accordingly over time. This allows, in particular, future time windows to be taken into account, enabling improved adjustment of the time window, which leads to improved operation of the interference-resistant network.

[0022] It has proven advantageous to generate the information message using a machine learning algorithm on the electronic computer. In other words, the machine learning algorithm can initially generate an initial information message and transmit it to the end device. Over time, particularly based on further received data packets, this information message, or a corresponding overview table, can be reliably and continuously updated, especially via a machine learning algorithm. The transmitted data packets thus serve as training data to reliably determine at least one, or even multiple, time windows when transmission is appropriate.

[0023] It has also proven advantageous to implement the learning algorithm as a neural network. In particular, this allows for the simple yet reliable generation of the information message and, especially, the summary table. Here and in the following, an artificial neural network can be understood as software code stored on a computer-readable storage medium that represents one or more interconnected artificial neurons or can replicate their function. The software code can also contain multiple software code components, which may, for example, have different functions.In particular, an artificial neural network can implement a nonlinear model or a nonlinear algorithm that maps an input to an output, where the input is given by an input feature vector or an input sequence and the output can include, for example, a category output for a classification task, one or more predicated values, or a predicated sequence.

[0024] Furthermore, it has proven advantageous to deploy the neural network as a long-short-term memory (LST) algorithm. The LST is an artificial intelligence with an input gate, a forget gate, and an output gate. In contrast to conventional recurrent neural networks, this allows for a kind of memory of past experiences. Specifically, it provides a long-lasting "short-term memory" because the network's fundamental behavior is encoded in its weights. This results in improved operation of the noise-resistant network.

[0025] It is also advantageous if a communication device of the terminal is deactivated depending on the specific time window. In other words, the terminal's modem is only activated when it corresponds to the relevant time window in the overview table. At other times, the modem is deactivated. This is particularly beneficial because modems consume a lot of energy and terminals, for example, only have batteries that need to be replaced, thus reliably extending the terminal's operating time.

[0026] In a further advantageous embodiment, it is provided that a further multitude of additional data packets are transmitted from the electronic computing device to the terminal device via the interference-resistant network, wherein each additional data packet is assigned a respective additional timestamp and a respective additional packet number by means of the electronic computing device, wherein at least a portion of the multitude of transmitted additional data packets is received by means of the terminal device, wherein unreceived additional data packets are determined by means of the terminal device based on the respective additional packet number, and wherein at least one further time window in which the unreceived additional data packets are sent is determined by means of the terminal device based on the additional timestamps of the received additional data packets.In other words, it is also possible to determine when a communication time or communication channel from the electronic computing device to the terminal device could be made available. This allows the electronic computing device to know when to send these data packets without any loss. Furthermore, this enables the terminal device to open a receiving channel only at these times, thereby saving electrical energy.

[0027] However, it is preferred that, since the end device, for example, does not have sufficient computing resources available, the determination of such an overview table is also carried out by the electronic computing device. Therefore, it is specifically intended that only the relevant additional time windows are transmitted to the electronic computing device, which can then create a corresponding overview table based on these additional time windows.

[0028] It is also advantageous if at least one additional time window is transmitted from the terminal device to the electronic computing unit, and if the electronic computing unit then generates a further overview table based on this information. As already mentioned, this additional overview table can then be used to determine when the electronic computing unit sends data packets to the terminal device. Furthermore, this additional overview table can also be transmitted to the terminal device, so that the terminal device knows when to activate the corresponding receiving channel.

[0029] It can also be stipulated that a data packet from the electronic computing device is only transmitted to the terminal device within the determined time window specified in the overview table. This prevents data packets from being lost between the electronic computing device and the terminal device. In particular, it allows the system to determine when the IoT device needs to activate its modem to receive incoming messages. This process is similar to the prediction process for the transmitting device. In this case, the data packets sent from the electronic computing device to the terminal device are also provided with a timestamp indicating the transmission time and an incrementing packet counter.Similar to the process on the end device's side, the terminal device now creates "window maps" from the incoming data packets. These window maps are no longer processed on the devices themselves, but are transmitted to the terminal device. They also contain the identifier of the received packets and the time they were received. The terminal device now knows when which packets reached the terminal device and when they were received.

[0030] It is further advantageous if, in addition, a transmission delay is determined by the electronic computing device based on the timestamp. In particular, this allows predictions to be made about the transmission delay, especially from the terminal device to the electronic computing device. The electronic computing device uses the transmission and arrival times to estimate how long the message takes to travel from the terminal device to the electronic computing device. Similarly, the terminal device reports back how long various messages took to reach the terminal device from the electronic computing device. Here again, machine learning algorithms, such as long-short-term memory models, can be used, especially as with the corresponding connection windows, to be trained on the time series of delays at the electronic computing device.The predicted delay profile is then transmitted to the end device, which can in turn use the table to estimate the future time until the message reaches the network edge or the electronic computing unit. This is necessary or useful, for example, for decision-making regarding the migration of time-critical tasks to the network edge or the electronic computing unit.

[0031] The presented method is, in particular, at least in part, a computer-implemented method. Therefore, a further aspect of the invention relates to a computer program product with program code means which, when executed by the electronic computing device, cause the electronic computing device to carry out a method according to the preceding aspect.

[0032] Furthermore, the invention therefore also relates to a computer-readable storage medium with at least the computer program product according to the preceding aspect.

[0033] A further aspect of the invention relates to a network system for a fault-resistant network, comprising at least one terminal device and an electronic computing unit, wherein the network system is configured to carry out a method according to the preceding aspect. In particular, the method is carried out by means of the network system.

[0034] Advantageous embodiments of the process are to be regarded as advantageous embodiments of the computer program product, the computer-readable storage medium, and the network system. The network system possesses tangible features to enable the execution of the corresponding process steps.

[0035] A computing unit / electronic computing device can be understood, in particular, as a data processing device containing a processing circuit. The computing unit can therefore process data to perform arithmetic operations. This may also include operations to perform indexed access to a data structure, such as a lookup table (LUT).

[0036] The computing unit may, in particular, contain one or more computers, one or more microcontrollers, and / or one or more integrated circuits, for example, one or more application-specific integrated circuits (ASICs), one or more field-programmable gate arrays (FPGAs), and / or one or more systems on a chip (SoCs). The computing unit may also contain one or more processors, for example, one or more microprocessors, one or more central processing units (CPUs), one or more graphics processing units (GPUs), and / or one or more signal processors, in particular one or more digital signal processors (DSPs). The computing unit may also include a physical or virtual array of computers or other units of the aforementioned type.

[0037] In various embodiments, the computing unit includes one or more hardware and / or software interfaces and / or one or more storage units.

[0038] A storage unit can be volatile data storage, for example as dynamic random access memory (DRAM) or static random access memory (SRAM), or as non-volatile data storage, for example as read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or flash EEPROM, ferroelectric random access memory (FRAM), or magnetoresistive random access memory.It can be designed as MRAM (magnetoresistive random access memory) or as phase-change random access memory, PCRAM (phase-change random access memory).

[0039] For use cases or application situations that may arise in a method according to the invention and that are not explicitly described herein, it may be provided that, according to the method, an error message and / or a request for user feedback is issued and / or a default setting and / or a predetermined initial state is set.

[0040] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0041] Further features and combinations of features of the invention will become apparent from the figures and their description, as well as from the claims. In particular, further embodiments of the invention need not necessarily include all features of any one of the claims. Further embodiments of the invention may have features or combinations of features not mentioned in the claims.

[0042] This shows: Fig. 1 a schematic block diagram of an embodiment of a network system; Fig. 2 a schematic block diagram for determining a time window; Fig. 3 another schematic block diagram according to an embodiment of a network system; and Fig. 4 yet another schematic block diagram according to an embodiment of a network system.

[0043] The invention is explained in more detail below with reference to specific embodiments and associated schematic drawings. In the figures, identical or functionally equivalent elements may be designated with the same reference numerals. The description of identical or functionally equivalent elements is not necessarily repeated with respect to different figures.

[0044] Fig. 1 Figure 1 shows a schematic block diagram according to an embodiment of a network system 10. In the present embodiment, the network system 10 comprises at least one terminal device 12 and an electronic computing device 14. The terminal device 12 is specifically designed as a so-called IoT (Internet of Things) device. In particular, the terminal device 12 can be designed as a so-called Industrial IoT device. The electronic computing device 14 is specifically provided as a so-called Edge Node.

[0045] The network system 10 is specifically designed to operate a fault-resistant network 16. According to one embodiment of the present method, a large number of data packets 18 are transmitted from the terminal device 12 to the electronic computing device 14 via the fault-resistant network 16, with each data packet 18 being assigned a respective timestamp 20 and a respective packet number 22. The electronic computing device 14 receives at least a portion of the large number of transmitted data packets 18 and identifies unreceived data packets 18 based on their respective packet numbers 22. Then, at least one time window 24 is determined ( Fig. 2 ), in which the unreceived data packets 18 were sent, based on the timestamps 20 of the received data packets 18 by means of the electronic computing device 14. An information message 26 with at least one determined time window 24 is then generated by means of the electronic computing device 14 and the information message 26 is transmitted to the terminal device 12 for future transmission of data packets 18 via the interference-resistant network 16 by means of the electronic computing device 14.

[0046] In particular, it may be provided that data packets 18 are transmitted from the terminal device 12 to the electronic computing device 14 within the determined time window 24. Furthermore, the information message 26 may include a summary table 30 ( Fig. 3 ) over which at least one determined time window 24 is transmitted. Furthermore, it can be provided that the electronic computing device 14 continuously determines time windows 24 and the information message 26 is continuously adapted and transmitted to the terminal device 12.

[0047] Furthermore, it may be provided that the information message 26 is generated by means of a machine learning algorithm of the electronic computing device 14. The machine learning algorithm may, in particular, be provided as a neural network 28. Furthermore, it may be provided that the neural network 28 is provided as a long-short-term memory algorithm. It may also be provided that, depending on the specific time window 24, a communication device of the terminal device 12 is deactivated.

[0048] In particular, this shows that Fig. 1 The restricted terminal 12 and the service instance, in particular the electronic computing device 14, are located at the network edge. These restricted terminals 12 are connected to the electronic computing device 14 via the interference-resistant network 16, which can be considered a black box. This connection is highly intermittent, and at no time can a direct connection exist between the restricted terminal 12 and the network edge, especially the electronic computing device 14, particularly in so-called delay- and interruption-tolerant networks. Due to the black-box nature of the interference-resistant network 16, the predictive approach is directly compatible with almost all types of network infrastructures.

[0049] In this approach, the terminal 12 uses lookup tables created by the electronic computing unit 14, particularly with the aid of deep learning models. Specifically, these are created using LSTM-based methods to extract recurring patterns, such as satellites passing by at the same time or trains arriving at approximately the same time according to timetables. These lookup tables contain the next estimated start times of connection windows and their durations. This predictive approach is fully compatible with limited field devices because essentially all the computing work is performed in the data centers, particularly the electronic computing unit 14, at the network edge or core.

[0050] The Fig. 1 This shows in particular the terminal device 12, located remote from the electronic computing device 14, with the intermittent network, for example data mules, subcum, disaster relief network, or a combination of these networks, with the electronic computing device 14. As already mentioned, the interference-resistant network 16 is treated as a black box.

[0051] For the sake of simplicity, only one terminal device 12 is shown in the present embodiment. It is self-evident that further terminal devices 12 can also be connected to the electronic computing unit 14 via the interference-resistant network 16.

[0052] In a first step, the terminal device 12 attempts to unload data packets 18 at regular intervals via the interference-resistant network 16 and therefore sends the corresponding data packets 18 to the electronic computing device 14. These data packets 18 are additionally provided with the timestamp 20, which describes in particular the local transmission time of the data packet 18. Furthermore, if not already present, an incremental packet number 22 is appended. This packet number 22 is incremented for each data packet 18 that is sent from the terminal device 12 to the electronic computing device 14.

[0053] Fig. 2 This shows a schematic view of how to generate time window 24. In particular, the upper part of the Fig. 2 It has been shown that, for example, twelve data packets 18 are sent from the terminal device 12 to the electronic computing device 14 via the interference-resistant network 16. In the present embodiment, it can be seen, for example, that data packets 3, 4, 5, and 6 were not received. This allows a time window 24 to be generated. In particular, since the connection between the terminal device 12 and the electronic computing device 14 exists only sporadically, some data packets 18 will arrive, while others will not. All incoming packet numbers 22 are stored, and on the electronic computing device 14, the so-called window map is created. This map represents all intervals in which the terminal device 12 was able to successfully send data packets 18 to the electronic computing device 14.Particularly due to the often asynchronous nature of interruption-tolerant networks, a connection from the terminal device 12 to the electronic computing device 14 does not automatically imply a corresponding connection from the electronic computing device 14 to the terminal device 12.

[0054] In the present embodiment, it is shown, for example, that the data packets 1, 2, 3, 7, 8, 9, 10, 11 and 12 arrive at the electronic computing device 14.

[0055] Fig. 3 Figure 1 shows another schematic block diagram according to an embodiment of the network system 10. It is shown in particular that the neural network 28, especially in this case LSTM-based, can be trained on the multitude of data packets 18 to predict the next offline and online windows. For this purpose, it can be provided that different neural networks 28 can also be used. These predicted time windows 24 are then stored in a lookup table, which is transmitted to the terminal device 12. The terminal device 12 can now use this lookup table to unload data only when there is a high probability that the network service contains these data packets 18.

[0056] Fig. 4 Figure 10 shows another schematic view of an embodiment of a network system. According to the Fig. 4 It is shown in particular that a further multitude of data packets are transmitted from the electronic computing device 14 to the terminal device 12 via the interference-resistant network 16, wherein each further data packet is assigned a respective further timestamp and a respective further packet number by means of the electronic computing device 14, wherein at least a part of the further multitude of transmitted further data packets is received by means of the terminal device 12, wherein unreceived further data packets are determined by means of the terminal device 12 on the basis of the respective further packet number, and wherein at least one further time window in which the unreceived further data packets were sent is determined by means of the terminal device 12 on the basis of the further timestamps of the received further data packets.

[0057] In particular, it may be provided that at least one further time window is transmitted from the terminal device 12 to the electronic computing device 14, and that the electronic computing device 14 generates a further overview table 32 based on this. Likewise, a data packet 18 can only be transmitted from the electronic computing device 14 to the terminal device 12 within the determined further time window according to the further overview table 32.

[0058] Furthermore, the Fig. 4 , that in addition, depending on the timestamp 22, a transmission delay 34 can be determined by the electronic computing device 14.

[0059] In particular, from the perspective of terminal device 12, it is thus possible to determine when to activate a corresponding modem to wait for incoming data packets 18. This process is similar to the prediction process for the sending device, especially from terminal device 12 to the electronic computing device 14. In this case, the data packets 18 sent from the electronic computing device 14 to terminal device 12 are also provided with an additional timestamp for the transmission time and an additional incrementing packet counter. Similar to the process on the electronic computing device 14 side, terminal device 12 also creates the additional time window from the incoming data packets 18, which are then no longer processed on the restricted terminal device 12 itself, but are transmitted to the network edge, specifically to the electronic computing device 14.Furthermore, these cards contain the identifier of the received data packets 18 and the time at which the data packets 18 were received. The electronic computing device 14 now knows at what time which packets reached the terminal device 12 and when they were received.

[0060] At this point, two additional LSTM-based neural networks 28 are trained at the network edge, specifically at the electronic computing device 14, to predict successful reception and transmission times, particularly from the electronic computing device 14 to the terminal device 12. The transmission network remains at the network edge and is used to predict when new data packets 18 can be sent to the terminal device 12. The receiving network, in turn, creates a lookup table, specifically the second overview table 32, which is transmitted to the terminal device 12. This table can then be used by the terminal device 12 to determine probable arrival times of future data packets 18 and, ideally, to save energy.

[0061] Furthermore, predictions about the transmission delay 34 can be made in a similar manner. The electronic computing device 14 uses the transmission and arrival times to estimate how long the data packet 18 takes to travel from the terminal device 12 to the electronic computing device 14. Similarly, the terminal device 12 reports back how long various data packets 18 took from the electronic computing device 14 to reach the terminal device 14. Here, too, two LSTM-based models, as with the corresponding connection windows, are trained on the time series of delays at the electronic computing device 14. Subsequently, the predicted course of the delays, especially from the terminal device 12 to the electronic computing device 14, is transmitted to the terminal device 12, which can use this table to estimate the future time until the data packet 18 reaches the electronic computing device 14.This is necessary or useful, for example, for decision-making when transferring time-critical tasks to the electronic computing facility 14.

[0062] In particular, a form of time synchronization between the electronic computing device 14 and the terminal device 12 is required for estimating the delay.

[0063] It should be mentioned here that data packets 18 are not sent solely according to these estimates for connection windows and delays. Otherwise, the network system 10 would be unable to adapt to new online windows that appear during predicted "offline phases." To further explore potential new connection windows, a kind of "exploration" must be maintained, which can be done, for example, by sending data packets 18 at random times, even if these times were predicted to be offline. Therefore, there is also a trade-off between exploration and packet loss rates.

[0064] Estimating network delays and utilization using neural networks 28, or more precisely, long-short-term memory-based networks, is already a well-established field. However, efficient data offloading with connection and delay prediction, especially with intermittent networks, is still relatively unknown. In particular, where other solutions use specialized routing algorithms or require knowledge of the network topology, external data sources are needed, whereas the solution presented here requires no knowledge of the network's internal workings or any external information for the prediction process. Furthermore, the corresponding algorithm does not need to be added or modified on the restricted terminal device 12 and the electronic computing device 14. All network solutions within the interference-resistant network 16 can remain unchanged.Furthermore, no power-intensive training or inference from the neural networks 28 is required on the limited terminal device 12.

[0065] In summary, it is thus possible to achieve bidirectional prediction of connection windows / delays in the interference-resistant networks 16 or other emitting networks for the terminal device 12 using long-short-term memory-based models without prior knowledge of the network's internal processes. Furthermore, this enables energy- and resource-efficient prediction of connection windows / delays, for example, for task allocation.

Claims

1. A method for operating a fault-resistant network (16) between an electronic computing device (14) and an end device (12) using a network system (10), comprising the steps of: - transmitting a plurality of data packets (18) from the end device (12) to the electronic computing device (14) via the fault-resistant network (16), whereby each data packet (18) is assigned a respective timestamp (20) and a respective packet number (22) by means of the end device (12); - receiving at least a part of the plurality of transmitted data packets (18) by means of the electronic computing device (14); - determining non-received data packets (18) based on the respective packet number (22) by means of the electronic computing device (14);- Determining at least one time window (24) in which the unreceived data packets (18) were sent, based on the timestamps (20) of the received data packets (18) using the electronic computing device (14); - Generating an information message (26) with the at least one determined time window (24) using the electronic computing device (14); and - Transmitting the information message (26) to the terminal device (12) for future transmission of data packets (18) over the interference-resistant network (16) using the electronic computing device (14).

2. Method according to claim 1, characterized by the fact that In the future, data packets (18) will be transmitted from the terminal device (12) to the electronic computing device (14) within the determined time window (24).

3. Method according to claim 1 or 2, characterized by the fact that the information message (26) is transmitted with an overview table (30) about at least one determined time window (24).

4. Method according to any one of the preceding claims, characterized by the fact that The electronic computing device (14) continuously determines time windows (24) and the information message (26) is continuously adapted and transmitted to the terminal device (12).

5. Method according to any one of the preceding claims, characterized by the fact that the information message (26) is generated by means of a learning algorithm of the electronic computing device (14).

6. Method according to claim 5, characterized by the fact that the learning algorithm is provided as a neural network (28).

7. Method according to claim 6, characterized by the fact that the neural network (28) is provided as a long-short-term memory algorithm.

8. Method according to any one of the preceding claims, characterized by the fact that Depending on the specific time window (24), a communication device of the terminal device (12) is deactivated.

9. Method according to any one of the preceding claims, characterized by the fact that a further multitude of additional data packets is transmitted from the electronic computing device (14) to the terminal device (12) via the interference-resistant network (16), wherein each additional data packet is assigned a respective additional timestamp and a respective additional packet number by means of the electronic computing device (14), wherein at least a part of the further multitude of transmitted additional data packets is received by means of the terminal device (12), wherein unreceived additional data packets are determined by means of the terminal device (12) based on the respective additional packet number, and wherein at least one further time window in which the unreceived additional data packets were sent is determined by means of the terminal device (12) based on the additional timestamps of the received additional data packets.

10. Method according to claim 9, characterized by the fact thatthat at least one further time window is transmitted from the terminal device (12) to the electronic computing device (14) and on the basis of which another overview table (32) is generated by the electronic computing device (14).

11. Method according to claim 10, characterized by the fact that a further data packet from the electronic computing device (14) is only transmitted to the terminal device (12) within the further time window determined according to the further overview table (32).

12. Method according to any one of the preceding claims, characterized by the fact that Additionally, depending on the timestamp (20), a transmission delay (34) is determined by the electronic computing device (14).

13. Computer program product comprising program code means which cause an electronic computing device (14) to perform a method according to one of claims 1 to 12 when the program code means are executed by the electronic computing device (14).

14. Computer-readable storage medium comprising at least one computer program product according to claim 13.

15. Network system (10) for a fault-resistant network (16), comprising at least one terminal device (12) and an electronic computing device (14), wherein the network system (10) is configured to perform a method according to one of claims 1 to 12.

Images